The Possibility of Self-Government.
To speak meaningfully of an organization or political community as self-governing requires that it have (or implies that it has) the property of having, knowing, and consciously pursuing its own "will." As an abstract proposition, this claim is not especially controversial, but exactly how the standard it sets might actually be met remains a matter of deep dispute. I will address a particularly strong form of skepticism about the possibility of satisfying this requirement. This is the claim made by certain social choice theorists that it is not merely difficult, but logically incoherent, to believe there are social procedures under which collectivities can learn their own will. I hope to show that, whatever other problems theories of the "social will" face, this particularly strong objection can be answered, and in a way that opens up fruitful avenues for further inquiry. I begin by briefly setting out the objection that I aim to defuse.
THE SOCIAL CHOICE OBJECTION
Starting with Kenneth Arrow's ( 1963) seminal monograph, social choice theorists have sought to identify procedures under which a collectivity might identify the preferences that could reasonably count as its own. Taking for granted relatively uncontroversial liberal democratic assumptions, the social choice theorists asked a simple question: How can one devise procedures to "process" or "aggregate" the impartially weighted preferences of individual members of a collectivity into a preference that can stand for the collectivity as a whole? The answer Arrow and his successors repeatedly gave was that the only procedures capable of identifying a coherent and unambiguous collective preference or "will" involve either deferring to the judgment of a dictator or narrowing in advance the range of preferences that citizens are permitted to express. Neither of these ways of generating a coherent collective preference seems to satisfy our intuitive sense of what it means for something to have its own will. No one would be satisfied by the suggestion that a society whose decisions are determined for it by a dictator is genuinely following its own will. Similarly, to impose a blanket prohibition on society choosing a certain preference ordering, even though all its members would otherwise do so, may also seem inconsistent with the notion of society following its own will. Arrow's terminology is telling here: To violate this second proviso is to violate a norm of "citizen sovereignty."
In the years since Arrow's initial contribution, social choice theorists have both confirmed and deepened the general point (McKelvey 1979; Plott 1976; Riker 1982; Schofield 1985; Sen 1970). As a result, they have been increasingly confident in their claims about the implications of social choice theory. It is now standard for these theorists to claim that the impossibility theorems, because they establish there can be "no such thing as the will of the people" (Maclean 1991, 509), mandate a certain kind of methodological individualism. Willing, preferring, and choosing are not possible predicates of societies, and to think otherwise is to engage in a primitive kind of anthropomorphism (Riker 1982, 18; Schelling 1984, 93). The comments of Charles Plott (1976, 525) are representative:
The concept of social preference involves an illegitimate transfer of the properties of an individual to the properties of a collection of individuals.... The Arrow theorem demonstrates that the concept of social preference involves the classic fallacy of composition, and it is shocking only because the thoughts of social philosophers from which we have developed our intuitions about such matters are subject to the same fallacy.
I seek to refute this kind of claim and to show how having a will might be a possible property of both societies and individuals. Before I tackle these claims, I want to set the technical issues raised by social choice theory within the broader context of discourse about self-government. This is essential if we are to understand the nature of the difficulty identified by social choice theory and its relation to the larger theoretical issues that it often claims to resolve.
SELF-GOVERNMENT: SOME PRELIMINARIES
Very roughly, to say that a political community or organization is self-governing is to say that in some sense or senses the actions taken or controls imposed by its governing institutions can be thought of as originating from within that community or organization. There is no essential or canonical specification of the senses in which self-government demands that authoritative action be internally "sourced" in this way. To set the stage for my argument, however, I will identify three salient ways in which one might claim that authoritative action originates from within a self-governing political community or organization. I will then use these three senses of self-government to make several preliminary points about self-government, democracy, and social choice theory.
A first and relatively undemanding sense in which a group can be self-governing operates at a basic institutional level. A society can be self-governing simply to the extent that the apparatus of government and administration is operated exclusively by officials and participants drawn from within, and authorized by, the citizen body, not by outsiders. In this institutional sense, imperial domination is the antithesis of self-government.
A second and more demanding interpretation of the idea that public action originates from within makes use of our organizing idea of a social will. In this view of self-government, public decisions must be plausibly understood by members of the collectivity as reflecting, expressing, or revealing a will that is authentically their own, and there must at least be social consensus on procedures for determining or verifying the content of this will, such that one can in principle assess the extent to which public action fulfills or deviates from it. Self-government in this sense requires that the political community itself be the subject of political action at least in the following minimal way: Both governors and governed must recognize accepted "will-revealing" procedures and view them as communicating instructions, which governing agencies are then expected to execute as a matter of course. If this condition were satisfied, we could meaningfully speak of a society or organization being the author of its own fate, at least insofar as the actions it takes successfully fulfill a will that has been appropriately certified as its own. It is the possibility of ever satisfying this condition that social choice theory calls into question.
A third and still more stringent sense in which public action might have an internal source results in a highly moralized (and I believe marginal) conception of self-government. In this account, society is self-governing to the extent that its government acts only in the light of a plausible conception of citizens' common good. The idea is that officials take the whole community's welfare (the "public interest") to be the main object of public action, and they appeal to a plausible account of this common good as the grounds for public action. In this sense, a community is self-governing insofar as public action is oriented only to those goals and ends identifiable as proper, rather than alien, to the whole community. This rules out justifications for public action that either appeal to goals and values that go beyond the welfare of the relevant community (the progress of mankind in general, fulfilling God's providential design, and so on) or privilege the interests of narrower sectional or partisan groups wi thin the community. For instance, juries might seem to be self-governing in a variety of other senses (e.g., juries do act in accordance with their will), but they are not self-governing in this third sense, for a jury does not derive justification for a verdict on a particular occasion from a set of claims about its own well-being but, rather, looks to something beyond itself--the fate of a third party.
Similarly, a society whose ruling class justifies the exploitation of the rest of the population by reference to its own partisan interests, rather than to those of the whole society impartially considered, is not self-governing in this third sense. In such a case, justifications for public action are not grounded in claims about the best interests of society as a whole but in claims about the interests of a partisan subset of the community. Here, it seems more accurate to say that one social group is simply subjugated to another in that the former's interests are not given due weight in the (purported) justifications for public action offered by the ruling class. One can reasonably claim that this sort of subjugation is inconsistent with at least ideal forms of self-government. Certainly, the history of modern concepts of self-government, from Rousseau on, is replete with examples of theorists and commentators who have understood the concept in this way. 
To set the stage for my argument, I now make several connected observations about these three senses of self-government.
Self-Government and Democracy
The concept of self-government has an elective affinity with the idea of democratic rule. But if democracy requires simply that all citizens in a given state (have a right to) participate in the political process, then strictly speaking it is not true (1) that any of these three senses of self-government requires democracy or (2) that democracy entails self-government in any of the three senses. Hobbes's political theory illustrates the first point: Citizens can will their collective security and self-preservation by authorizing highly undemocratic forms of government; even a monarchical Leviathan instantiates Hobbesian self-government.
The second point is illustrated by those empirical theories of democracy inspired by the work of Joseph Schumpeter (1942) and his antecedents (Michels 1916). These modern theorists of democracy regarded the notion of self-government as obsolete and prescientific, in part because in their view it tends to obscure the systematically oligarchical reality of political organizations. Despite this, they believed that a theoretically useful theory of democracy could be retrieved from the conceptual swamp left behind by the philosophers of self-government.  The findings of social choice theory tend to reinforce the intuitions of those in the Schumpeterian tradition. In this view, of which William Riker (1982) was the most vociferous proponent, the social choice objection demonstrates that the only viable conceptions of democracy available to us are ones that drop the idea that democratic votes express a popular will. We are left with a much less ambitious--in Riker's terms, "liberal"--conception of democracy. It demands merely that all citizens (have the right to) participate in a set of political processes that provide some kind of external check on the activities of oligarchical political elites. The ability of democratic political systems to institutionalize this sort of check does not, however, imply that when the check is invoked, "society" as a whole is concertedly acting in accordance with something called "its own will" (Riker 1982, 243-5). Rather, the democratic system can be thought of as a rather crude, although at least inclusive, arbitration device. According to Riker and his followers, social choice theory provides a cast-iron justification for severing the connection between democracy and the problematic idea of self-government. This conclusion is apt to seem shocking: If authentic democracy requires anything, it is surely that public action track the popular will. But it is just this conception of democratic self-government that social choice theory calls into question. In order to retrieve that conce ption, we must defuse the social choice objection and show how there can in principle be such a thing as the popular will.
Self-Government and the Common Good
Self-government in the third and most demanding sense shades into, and becomes difficult to distinguish from, some more general account of political justice and the social good. For this reason, we should regard this third sense as lying at the very periphery, rather than near the core, of the concept of self-government. This is supported by the intuitively plausible reflection that a self-governing society need not be an especially good society.  The first two senses of self-government preserve that intuition, but the third sense overturns it. A society could be governed only by its own members (first sense), and it could know and pursue its own will (second sense), but from neither supposition does it automatically follow that this society acts in accordance with its common good. The social will might be misguided or even evil, and those citizens who wield power might be entirely corrupt. The third sense, however, precludes these possibilities, because when policy runs counter to the correct appreciation of the common good and its imperatives, the relevant society is not fully self-governing. But this seems far too strong. It might make sense to say that such a society falls short of the moral ideal of a perfectly self-governing community, but it is implausible to say that it is violating conditions that are either strictly necessary or sufficient for us to be able to describe it as genuinely self-governing.
The Central Sense of Self-Government
If the third sense seems too strong, the first seems too weak by itself to count as a sufficient condition for self-government. Yet, the first sense does appear to embody a necessary condition for regarding a community or organization as authentically self-governing. Although the absence of (say) imperial domination does not automatically guarantee that a society is self-governing (there have been plenty of nonimperial autocracies), it is clear enough that any form of imperial domination is strictly incompatible with self-government. By contrast, the second sense looks very much like a sufficient condition for self-government: If we can show that a society is acting upon a will that it knows to be its own, it is unlikely that we need to establish anything more in order to convince ourselves that this society is self-governing. There might be a further question about whether this society is good, but we have already seen that a self-governing society need not be good: The mere fact that a certain society is pu rsuing policies inconsistent with the common good or social justice, or that are in some other way ethically problematic, does not automatically establish that it is not self-governing. As long as public policy fulfills the social will (for good or ill), that seems sufficient to show that this society is self-governing.
The Independence of the Second and Third Senses
These observations suggest not only that the second sense of self-government is the central one but also that it is largely independent of the third. This is a pivotal point in my argument. One of the main claims I advance is that many of the difficulties involved in the notion of self-government can be removed by maintaining a clear distinction between the second and third senses. That is, we should separate the question of how society can determine its own will from that of how we might discern the public interest or society's common good. Both classical and contemporary discussions of self-government have systematically confused these two issues. Two characteristic forms of this confusion have particularly plagued these discussions.
Confusion 1. Sometimes it is assumed that there is ultimately no way to specify the common good apart from somehow amalgamating citizens' judgments on some occasion about what the common good requires. The effect of this move is to collapse the third sense of self-government into the second: In this view, the notion of the common good involved in the third sense is identified with whatever society contingently nominates as its common good through the kind of "will-revealing procedures" envisaged in the second sense. Social choice theorists are particularly prone to making this move, mainly because they subscribe to an attenuated version of utilitarianism that views the public interest or common good as an aggregative function of the contingently expressed preferences of a society's constituent individuals. Therefore, they are tempted to claim that answers to questions about society's common good or the public interest stand or fall on the results of voting procedures that aggregate citizens' various views on the subject. Consider in this regard Riker's (1982, 137) claim that social choice theory "even leads us to suspect that no such thing as the 'public interest' exists, aside from the subjective (and hence dubious) claims of self-proclaimed saviors" (see also Plott 1976, 526-7). Riker seems to think this follows from the fact that the impossibility theorems show that outcomes of votes will systematically be ambiguous or meaningless.
But this inference is surely invalid. We would not normally conclude from the fact that individuals make contradictory or ambiguous claims about their own interests that there is no longer any basis on which to make credible assessments of their best interests. Still less does it show there is "no such thing" as an individual's best interests. Perhaps the individual in question is simply confused or ill informed. The same point applies with equal force in the case of claims about the best interests of a collectivity (i.e., the common good). The fact that citizens' amalgamated judgments about their common good are (even systematically) ambiguous or incoherent is not a reason to suppose that no such thing as the common good or public interest exists or can be determined. The salient question in this context is: What is the common good, and what does it require under circumstances A, B, C? But a claim about the ability or inability of a particular agent or agency to enunciate a clear, unambiguous answer to this question does not amount to any sort of answer to that question itself. We would at the very least need some additional argument to show that this particular agency's opinion on the subject carries special authority, so that its failure (say) to arrive at a clear, coherent answer definitively establishes that the question itself just is finally unanswerable. After all, the fact that some agent manages to enunciate a coherent, unambiguous view of its own best interest does not in any way imply that that view is correct or even plausible.
To my mind, these reflections highlight the fundamental gap between two issues raised, respectively, by the second and third senses of self-government. The second sense raises the question of how we can establish that a particular judgment about the common good can genuinely be regarded as society's own will. The third sense raises the different issue of how we can determine a correct or plausible account of society's common good. Perhaps social choice theory puts in doubt the possibility of ever resolving the first issue, although that is the claim I seek to refute. But even if my argument fails, it still does not follow that social choice theory puts in doubt the possibility of resolving the second issue; determining the common good is a different problem, and there is little reason to believe the apparatus of social choice theory is equipped to address it. Note, finally, that one can admit all this without in the least casting doubt on the proviso that, properly understood, the common good is a function o f the welfare of the individuals who make up the political community. This unobjectionable proviso must be distinguished from the very different and more dubious claim that the common good is whatever citizens contingently judge it to be on some occasion.
Confusion 2. A second way in which the second and third senses of self-government are often conflated moves in just the opposite direction, by effectively collapsing the second sense into the third. This occurs when one succumbs to the temptation to stipulate that some putatively valid account of the common good (and its correct application to contingently arising social and political problems) embodies the "real will" of the people. On this sort of view, in order to disclose the people's will on some occasion, we need only grasp the correct account of their common good and determine what it recommends under the relevant circumstances. This proposal derives such plausibility as it has from the Socratic intuition that agents, to the extent they are rational, will their own good. When individuals or groups will what is counter to their best interests, they must have failed to understand or grasp what they "really" will. This assumption undoubtedly underlies a good deal of the mistrust of popular rule voiced ove r the centuries. If we suppose that the average citizen is not competent to determine the common good, then the "real will" of the people may be better served by taking power out of their hands and giving it to "the experts."
Some political theorists have tried to defend popular rule even within the terms of the Socratic assumption. Rousseau (according to some influential interpretations) is responsible for the most historically significant version of this argument, and for that reason I refer to it as the General Will argument. It confuses the social will and the common good in exactly the way I am trying to identify. In order to diagnose this second confusion and highlight its shortcomings, I will briefly discuss some salient features of the General Will argument.
THE GENERAL WILL ARGUMENT
The common good requires that we give fair consideration to the best interests of all citizens, and the General Will argument suggests that the citizenry in full assembly is the appropriate "expert" on the common good. To the extent that we entrust power to groups smaller than the whole citizenry, we increase the likelihood that government will unfairly privilege the interests of some partisan sector of the community. This argument need not imply that the whole citizenry is infallible, merely that a virtuous citizenry is more likely to identify correctly the general will than are any of the less inclusive alternative candidate groups. The latter are more likely to be unduly influenced by their narrow sectarian interests and so are likely to offer correspondingly less reliable guidance as to the content of the general will or common good.
Today, political theorists appeal to versions of this argument in order to respond to the difficulties raised by social choice theory. But if social choice theory calls into question the idea that we can identify a will that is society's own, it is not clear that the General Will argument is capable of supplying an adequate response. The reason is that this argument implausibly reduces the concept of the social will to that of the common good. Consider a contemporary variant of the General Will argument that has been defended by a number of theorists who sometimes describe themselves as "epistemic populists" (Cohen 1986; Coleman and Ferejohn 1986; Grofman and Feld 1988; Radcliff 1992).
Epistemic populists respond to social choice theory by defining the general will as an "independent standard of correct decisions--that is, an account of justice or of the common good that is independent of current consensus and the outcomes of votes" (Cohen 1986, 34). This is exactly the reduction of the social will to the common good to which I am objecting. Epistemic populists then require that voters express their sincere beliefs about the content of the general will in specific instances. Under this requirement, voters do not merely assert whatever unreflective preferences happen to pop into their heads, so epistemic populists believe that "public deliberation is guided by the principles that define... [the general] will" (p. 34). This transforms the context and purpose of voting. Citizens are now expected to vote on the basis of good faith attempts to discern the general will. Clearly, this obviates social choice theory's search for an all-purpose aggregation function that can reveal a social preferenc e in the absence of any restrictions on the preferences individuals are permitted to express. Moreover, under these conditions, the outcomes of votes can constitute compelling evidence about the content of the general will or common good, or so the argument goes. 
As will become clear, my own argument shares many features of this approach. But as it stands, this epistemic populist proposal does not satisfactorily explain why the outcomes of these sorts of votes should automatically count as the social will. Epistemic populists simply recruit their specially configured voting procedures to the project of determining what some antecedently specified and independently valid account of the common good requires under specific circumstances. Even if we concede that epistemic populist voting procedures do indeed provide reliable evidence about the common good and its imperatives, is it obvious that we establish the content of the popular will by applying them? Suppose a society conducts a vote along the lines recommended by the epistemic populist. Citizens now have good reasons to believe that the victorious policy conforms to, or is dictated by, the common good. But are they entitled to infer from this fact alone that the winning policy constitutes their own will? Only if t hey already know that the epistemic populists' voting procedures are an appropriate barometer of their collective will. Yet, the mere fact that those procedures are built around a plausible view of the common good does not establish that this is so: Perhaps there is a conflict between what the common good actually requires of a society and what that society might--on further reflection--choose to will for itself.
Suppose, for example, that putting a man on the moon is a complete waste of valuable social resources from the point of view of the common good: The money would be far better spent improving the condition of disadvantaged citizens. This fact alone hardly establishes that a society has collectively decided to stop funding moonshots. After all, members of society might choose to sacrifice their common good to the goal of putting a man on the moon. Perhaps they see themselves as responsible not just for their own common good but for the progress of mankind in general. In such cases of a conflict between the common good and the social will, the content of that will cannot simply be read off the best available account of society's common good, even if we determine the latter through some sort of voting procedure.
At best, then, epistemic populism justifies a kind of social decision procedure that can be expected to (1) generate meaningful results and (2) provide reliable guidance about what the common good requires on specific occasions. But these do not constitute answers to the question which I am considering and which social choice theory claims is unanswerable. That question is not: What is the common good, and what does it require? The question is: What is it that we, as a society, actually want to do? To be sure, this latter question requires that we identify some clear, unambiguous result, but it does not require that the result be mandated by the common good. Instead, it requires that in arriving at a certain result, society must "know" that it is being true to itself or must authentically express its own view (right or wrong) of what ought to happen. If we are to respond to the social choice objection, then we have to show how this can be so, and how citizens can come to know that this is so.
To the extent that epistemic populists and other proponents of the General Will argument offer any answer to this question, it comes in the form of the Socratic stipulation that the social will and the common good must ultimately be one and the same. Defining the general will as the common good properly understood effects just this equation. But as an answer to our question about the possibility of self-government, this stipulation has at least two decisive drawbacks. First, notions of the common good or the public interest are not exactly unproblematic. If we are seeking to dissolve the ambiguities and problems of the notion of a social will, then an appeal for aid to the equally vexed notion of a "correct account of the common good" is like throwing out an anchor to save a drowning swimmer. There is every reason to demand that unravelling the mystery of the social will not await a solution to the riddle of the common good.
Second, stipulating an equivalence between the social will and the common good radically distorts the relation between the idea of self-government and the notion of freedom. Intuitively, the concept of self-government commands our moral attention in large part because it seems to be a condition for a certain kind of social freedom. A society that is not self-governing is in some way subjugated, enslaved, subordinated to alien forces, and so on, and in these senses is unfree. By contrast, a society that acts on a will that it knows to be its own is (to that extent) free in the sense relevant to the concept of self-government. But such a society is surely no less free in this sense if its will is misguided or runs counter to its common good. To establish that a society is free in this sense, we need only establish that social action is loyal to society's own view (right or wrong) of what ought to be done. If we follow the General Will argument, however, and stipulate that society's real will or general will is whatever the common good properly requires, then we completely invert these perfectly sound intuitions about freedom. If freedom is a function of acting in accordance with the social will, and the social will is really the common good, then whenever a society opts for a course of action inconsistent with the common good it is acting unfreely, because it is not (really) acting in accordance with its (real) will. On this view, assuming for the sake of argument that he understood the common good plausibly, Plato's Republic becomes the ideal of a free, self-governing political community, whereas the Athenian democracy that put Socrates to death becomes its antithesis. This is surely absurd, quite apart from Plato's idiosyncratic view of the common good. We may agree with Plato that Athens made a catastrophic mistake when it condemned Socrates, but it seems bizarre to claim that for this reason Athens was acting unfreely or was somehow subjugated to "alien forces" when it chose to convict Socrates. But by stipula ting an equivalence between the social will and the common good, proponents of the General Will argument are forced into this deeply counterintuitive account of social freedom.  The price of reducing the social will to the common good is the sacrifice of the concept of freedom implicit in our intuitive concept of self-government.
Ideally, of course, we might hope that members of a fully self-governing society both organize themselves so as to secure their common good and know that in doing so they are enacting a will that is authentically theirs. When both conditions are met, we can say that society is freely and willingly pursuing its true common good. But the means by which we determine that social practices and policies conform to the common good and that they fulfill society's will are, and should be, independent. If we conflate them in the manner of the General Will argument, we risk losing our grip on the form of social freedom that lends moral weight to the ideal of self-government.
The lesson, then, is that the second and third senses of self-government should be kept apart. It is a mistake to suppose that the absence of a social will implies that there is no way to determine the content of the common good (the first confusion), but it is also wrong to think that appealing directly to some conception of the common good and defining it as society's "real will" (the second confusion) allows us to supply a satisfactory answer to the social choice skeptic. Such a response requires a direct and free-standing account of how a political community might identify a will as its own. Below I attempt to show that the results of social choice theory at least do not shortcircuit this project.
THE ANALOGY BETWEEN INDIVIDUAL AND SOCIAL WILL
The concept of collective self-government suggests an analogy between individual agents and collectivities. According to this analogy, a collectivity can be an actor in its own right, capable of formulating and determining its own will, and thereby responsible for its own choices and self-determination. Social choice theorists typically claim that their results "demonstrate" that any such analogy "involves an illegitimate transfer of the properties of an individual to the properties of a collection of individuals" (Plott 1976, 525). The social choice theorems largely derive their shock value from their claim to expose such analogies as logically confused.
This claim is mistaken, however. Understanding why is the key to my strategy in this paper. To see this, however, it is first important to grasp the overall structure of the argument that persuades social choice theorists of this claim. This argument has four steps: (1) We can formulate a number of conditions that any social decision procedure capable of generating clear, unambiguous results would have to satisfy in order for us to certify its ability to enunciate a "will" that is authentically society's own. (2) It is logically impossible to satisfy all these conditions in a way that permits meaningful, coherent results to emerge. (3) Therefore, there is no possible decision procedure for identifying a social will since there is no way to satisfy the authenticity criteria and at the same time produce clear results. To say that no such procedure is logically possible is tantamount to showing that there is no such thing as the popular will. (4) Because we routinely assume that when individuals decide on a cou rse of action or preference, those decisions do or at least may reflect their own will, it follows that collectivities are to this extent fundamentally unlike individuals. They are not agents in the way that individuals are because it is meaningless to think of them as having their own will.
This argument is decisive only if we are sure that the social choice theorist's conditions are reasonable constraints on purportedly will-revealing procedures. If they are unreasonable or nonessential, then the argument collapses, because the violation of those conditions will no longer automatically imply that a coherent result does not authentically represent the social will. I suggest that the conditions imposed by social choice theory are not necessarily essential or even reasonable in the context of attempts to identify a social will. If so, then will-revealing procedures may legitimately violate these conditions, and the idea that society can identify a will of its own is not threatened by the results of social choice theory.
We need to understand, therefore, the conditions imposed by social choice theorists and the implicit expectations they make of reasonable will-revealing procedures. I will focus on five conditions.  The first embodies the expectation that these procedures generate transitive results:
T (Transitivity): For any alternatives x, y, and z, if x is preferred to y and y is preferred to z, then x is preferred to z.
For our purposes, T is the only relevant expectation of decisions. Preference orderings that violate T are said to be "cyclical." The problem with cyclical orderings is that they fail to communicate a clear, unambiguous preference. The relevance of this condition to the present argument is clear: If the public is to be in a position to communicate a will to governments expected to execute it, then acceptable will-revealing procedures must satisfy T. Otherwise, the procedures will enunciate cyclical judgments and thus fail to identify any clear instructions or decisions to the institutions authorized to execute the social will.
The remaining four conditions embody expectations of the amalgamation procedure that is supposed to process the inputs of each citizen into a final judgment representative of society's own will.
P (Weak Pareto Principle): For any alternatives x and y, if all individuals prefer x to y, then society prefers x to y.
D (Nondictatorship): There is no individual such that, for every set of individual orderings, for all alternatives x and y, if that individual prefers x to y, then society prefers x and y.
U (Unrestricted Domain): For any set of alternatives and any set of individuals, the domain of the social welfare function includes all orderings of alternatives by individuals.
I (Independence of Irrelevant Alternatives): For any alternatives x and y, if the preferences of all individuals as between x and y remain the same, then the preference of society as between x and y depends only on the preferences of individuals as between x and y.
In what follows, I will refer to this set of conditions on both decision and aggregation as the TUPID conditions.
The social choice theorems show that we cannot satisfy all the TUPID conditions at once: We must choose between T and one or other of the remaining four. This claim is the second of the four steps in the argument I set out above. But can this second step establish the later claims in the sequence? Can it show that there is no such thing as the popular will and that the analogy between individual and collective agency must be spurious?
To see that the answer is "no," consider the following two claims: (A1) Societies are potentially like individuals because they can have their own will, and (A2) societies are potentially like individuals because both are able to determine the content of their will in analogous ways.
Note that A1 does not imply A2. It is possible that both individuals and collectivities can have a will of their own, but that there is no resemblance whatsoever between the respective procedures under which these two kinds of agencies determine the content of their will. But A2 assumes A1: If we know that societies are able to determine their will in the same way that individuals do, then we presuppose that societies can have a will.
The difficulty facing the social choice argument is this. Social choice theorists claim to falsify Al (there is no such thing as the popular will), but to do so they deploy an argument compatible with the truth of A2. Their argument cannot decisively falsify Al, since it is presupposed by A2.
How is the social choice argument compatible with the truth of A2? The answer is that social choice theorists (rightly) acknowledge the possibility that the TUPID conditions might also apply to the case of individual decision making (Hurley 1989, 226f; May 1954; Roemer 1996, 25-6; Steedman and Krause 1986). The standard way of making the case points out that when individuals are called upon to express preference rankings over defined alternatives (say, candidates for a professorship), they very rarely have the luxury of being swayed by a single dimension. Rather, they must rank candidates relative to a range of emergent preferences along different dimensions: teaching ability, reputation, quality of written work, affability, compatibility of research interests with members of the department, and so on. Moreover, these dimensions will very often generate widely divergent rankings. In principle, an individual seeking to arrive at a final transitive ranking must find some way to process these emergent lower dim ensional rankings into an overall (all things considered) ranking, much as a society must find some way to amalgamate individuals' preferences into a social preference. If the TUPID conditions define reasonable will-revealing procedures in both these cases, then A2 has been conceded.
It is relatively easy to stipulate equivalent TUPID conditions for the case of intrapersonal deliberation. I will follow Hurley's (1989, 231) extremely helpful stipulation of TUPID analogues for individuals seeking a "coherence function" capable of processing several criterial rankings of alternatives into an "all-things-considered" ranking. Apart from T itself, which remains unchanged, Hurley amends the other four conditions as follows.
[P.sup.*] (Weak Pareto Principle): For all alternatives x and y, if all criteria rank x above y, then x ranks above y all things considered.
[D.sup.*] (Nondictatorship): A coherence function must not give so much weight to one criterion that it outweighs any criterion that conflicts with it under any circumstances. That is, it must not be the case that there is one criterion such that any one alternative's superiority over any other according to this criterion would always result in its superiority all things considered, regardless of how other criteria might rank those alternatives.
[U.sup.*] (Unrestricted Domain): For any set of alternatives and any set of criteria, the domain of the coherence function includes ordering of all alternatives by criteria.
[I.sup.*] (Independence of Irrelevant Alternatives): For all alternatives x and y, if the rankings of x and y by all criteria remain the same, then the ranking of x and y all things considered remains the same; that is, a deliberator's ranking of a pair of alternatives all things considered depends only on the ranking of those alternatives by all the relevant criteria, not on the rankings of other alternatives.
If it is plausible to claim that these conditions apply to individuals trying to process several emergent preferences over given alternatives, then it would seem to follow that transitive decisions can be reached only by violating one or other of the [TUPID.sup.*] conditions. Roemer (1996, 26) suggests that this conclusion is an "unexpected bonus" of Arrow's theorem. But it seems to me that it gives the game away: If these conditions apply in the individual case, then this undermines the ability of social choice theory to adjudicate the question of whether there is such a thing as the popular will. Once we allow this possibility, we concede the possible validity of A2, and we can no longer claim to be establishing that A1 is false.
To see that this is right, suppose it is plausible to depict the individual will along these lines. The violation of one or other of the TUPID conditions then becomes a generic condition of all aggregation that results in transitive outcomes, whether individual or collective. If we assume that this form of aggregation captures essential features of any will-revealing procedure (whether individual or social), then this would imply either (1) that neither individuals nor collectivities have a will or (2) that in both instances the violation of the TUPID conditions is a precondition for identifying such a will.  Clearly, neither claim is capable of establishing that the ability to will belongs to individuals but not to collectivities.
Alternatively, the social choice theorist could simply drop the idea that the TUPID conditions apply in the individual case and say that they are appropriate only in the collective case. This at least would allow one to claim that individuals can have wills but collectivities cannot. There are two difficulties, however. First, this cannot serve as an argument to the proposition that it is futile to model collective decision on the analogy of a single individual. The argument rests on the assumption that this analogy is misguided. This simply begs the question against the claim that A2 is false. Even if we certify independently that A2 is false, this still will not falsify A1, since A1 does not require A2. This helps us understand why the social choice framework is in no position to test the claim its practitioners often see it as proving: that the analogy between collective and individual agency is flawed because collectivities cannot be said to share the individual property of having a will. Plott's claim t hat Arrow's theorem "demonstrates" the fallacy of treating collectivities like individuals cannot be sustained.
Second, adopting this line of argument results in a crucial shift in the burden of proof. If one concedes that the TUPID conditions need not apply to individual will-revealing procedures, then the case for their salience in the collective case is weakened. If it is permissible for individuals to override one or other of the TUPID conditions when they determine the content of their will, then why should it be impermissible for a collectivity to do so? We have no choice but to take the individual will as the basic primitive unit of comparison--there are no other places to look for accounts of how the will operates. But if it turns out that in determining and pursuing their own wills individuals routinely and legitimately violate the [TUPID.sup.*] conditions, then the case for expecting collectivities to satisfy them might be undermined. I will now explore this possibility, although the findings are neither conclusive nor complete. 
RECONSTRUCTING THE INDIVIDUAL WILL
Thus far I have argued that in order to be self governing, societies and groups need some procedure for certifying that some policy fulfills or conforms to its own view of what should happen. I maintained that such a procedure must be independent of procedures designed to determine the content of the common good. Correctly discerning the common good is one thing, and fidelity to society's own judgment is another. The latter matters more in the context of self-government. To be self-governing, a group's will must authentically be its own. I will call this the requirement of reflexive authenticity.
Social choice theorists believe that Arrow's theorem undermines the possibility of such a will because they take the TUPID conditions to define a minimal, but essential, standard of reflexive authenticity for collective will-revealing procedures. This is clear from the justification social choice theorists normally offer for the critical P, I, and D conditions. They argue that when these conditions are violated, social choice is not determined by voters' own judgments but is instead an artifact of the aggregation method (Riker 1982, 117-9). This is a technical way of expressing the demand that procedures for determining social choice should not preempt citizens' own choices. P, D, and I comprise social choice theory's test of reflexive authenticity.
But do these conditions represent the only reasonable standard of reflexive authenticity in the context of self-government? Perhaps there exist other reasonable interpretations of reflexive authenticity that are inconsistent with P, D, and I. I have suggested that we might seek such alternative conceptions by questioning the claim that analogous conditions ([P.sup.*], [D.sup.*], and [I.sup.*]) are essential constraints on individual will-revealing procedures, although many theorists have considered this claim (Roemer 1996, 25-6; Steedman and Krause 1986, 207-10). But if it is wrong, understanding why might allow us to reconstruct a notion of a popular will that is immunized against the threat posed by Arrow's theorem. We might learn why violating one or other of these conditions is consistent with an appropriate test of reflexive authenticity, thus undermining the basis for the social choice objection to the idea of a social or popular will.
How might one call [P.sup.*], [D.sup.*], and [I.sup.*] into question? It is tempting to focus on [I.sup.*]. As is well known, [I.sup.*] prohibits (among other things) cardinal rankings of alternatives and thus rules out claims about the relative intensity of several emergent preferences. But surely individuals should be permitted to take such information into account when they deliberate about their choices? This is certainly right, but in the present context the objection is insufficient. Relaxing the requirement of ordinality is unlikely to help us reconstruct a viable account of will formation for two reasons.
First, intensity of preference may predict behavior, but the content of one's will is another matter. Prima facie, there is no reason to expect that the content of one's will ought to be considered a simple function of the intensity of preferences, or of behavior. Our intuitive concept of the will allows, for example, that it is often weak relative to intense desires, even that it can be "betrayed" by intense temptations and actions prompted by them. This suggests that even if we had a viable measure of intensity (itself doubtful), that variable would not predict, or stand in any determinate relation to, the content of one's judgments about one's will.
Second, we must remember that we are investigating the individual will not for its own sake but for the purposes of illuminating possible collective will-revealing procedures. Taking into consideration the intensity of citizens' preferences will only be useful in the latter context if we have some interpersonally applicable standard of intensity. But there is overwhelming agreement among economists and social choice theorists that interpersonal cardinal comparisons of utility are meaningless (indeed, many think the same is true of intrapersonal comparisons). This consensus might be misguided, but relaxing [I.sup.*]'s requirement of ordinality will help us reconstruct an account of the collective will only if we can defeat the argument that interpersonal comparisons of intensity are meaningless. This route seems unpromising, and it is certainly unlikely to persuade a social choice theorist.
In what follows, I sketch a more promising line of argument against P, D, and I. It hinges on the idea that practical deliberation about how to be true to oneself has a characteristic depth structure that is incompatible with the TUPID model of the will. I follow a number of philosophers in using the hierarchical metaphor of orders of desires, preferences, volitions, and reasons to explicate this important dimension of will formation. To elaborate the point, I make extensive reference to Frankfurt's (1988, 11-26) classic, "Freedom of the Will and the Concept of a Person." I use Frankfurt largely for reasons of convenience and ease of exposition; I do not necessarily endorse his views on the definition of persons or the subject of free will.
Consider the case of the tempted spouse, who finds herself contemplating infidelity because of three distinct emergent desires. First, her husband no longer satisfies her appetite for sexual experimentation and variety; sex within her marriage has become routine and has lost its earlier sparkle. Moreover, she is strongly attracted to a colleague at work, who has given strong indications of a reciprocal interest. Second, her husband cheated on her three years ago, and she has harbored a latent desire for revenge ever since. Third, the colleague is the tempted spouse's boss, and she has reason to believe that a sexual relationship with him will advance her career. Although the tempted spouse experienced emergent desires in the past (e.g., overwhelming sexual interest in her husband) that would rank fidelity over infidelity, no such desires are present now. Her emergent desires unanimously recommend infidelity, but the tempted spouse still might decide, in the face even of this unanimous recommendation, to refr ain. The feature in virtue of which this is possible is the capacity for forming what Frankfurt calls "second order volitions."
Frankfurt (1988, 16) claims that we do not merely experience emergent first-order desires of the sort that dominate the mind of the tempted spouse. We also have the ability to form second-order desires, that is, desires to have or not to have certain desires. A second-order desire becomes a second-order volition when a person desires that a certain (first-order) desire move, or not move, them to act. A second order volition is, therefore, a desire to be moved or not to be moved to action by a certain desire or set of desires.  By virtue of this capacity, the tempted spouse may choose not to identify her will with her emergent first-order desires. It is important to emphasize that this does not imply she will successfully battle these temptations; she may give in to them. The point, however, is that even if her unanimous first-order desires cause an act of infidelity, this action does not necessarily receive the blessing of her will. It is more accurate to say that temptation caused her to act against her will, which will be reflected in the self-recrimination, shame, and guilt likely to follow her capitulation.
What sort of reasons might agents have for countermanding or even disowning their first-order desires by refusing to identify them with their will? Very roughly, the most obvious can be collected under the general category of the personal ideal. A personal ideal here refers to any sustained intention, aspiration, hope, or plan to become a certain kind of person or to live up to a certain exemplar. The tempted spouse may refuse to identify her will with her emergent desires because she is antecedently committed to "being a good spouse" or "being a good Christian." If those ideals demand that agents not be moved by lust, vengeance, greed, and so on, then people who identify with those ideals may try to renounce the disqualified desires. One reason to affiliate such ideals with one's "will" is to encourage vigilance toward motivations that might deflect or undermine one's resolve to be a certain kind of person. By refusing to identify her first-order desires with her will, the tempted spouse is trying to do jus t this; and her will to realize whichever personal ideal is operative in her case is instantiated in the formation and pursuit of certain second-order volitions about emergent desires.
Frankfurt's account of second-order volitions allows one to distinguish between two different ways in which agents may fail to generate or be in a position to pursue their own will. First, they may fail to form second-order volitions at all; Frankfurt calls them "wantons." It is vital to note that wantons may have second-order desires, and they may have the capacity to deliberate. The distinctive feature of a wanton is "mindless indifference to the enterprise of evaluating his own desires and motives"; the wanton may experience the second-order desire to suppress his first-order desire for sex or drugs, but he is indifferent as to which should win out. It follows that a wanton could not be self-governing in Frankfurt's sense because he could not care about fulfilling any personal ideals of the sort sketched above. Second, as illustrated by the case of the tempted spouse of Frankfurt's (1988, 16-9) example of an unwilling addict, an agent may be capable of forming a will, but it does not follow that the will is effective in generating action.
The cornerstone of a Frankfurtian will-revealing procedure, then, is a higher order review of currently emergent first-order preferences. Agents seeking to arrive at an all-things-considered judgment as to the content of their will, given a variety of rankings based on emergent first-order desires, must ask a version of the following question: If I allow my will to be determined by emergent preferences of this particular character, will I be able to think of myself as living up to some personal ideal with which I identify? If the answer is "no," the agent may disregard the recommendation. There are two things to note about this test. First, it can legitimately claim to be one of reflexive authenticity: It reveals whether agents are being true to themselves by following one preference or another. Second, it does not evaluate emergent preferences by directly assessing their recommended rankings. Instead, it assesses the character of the values generating these rankings and their relation to some operative pers onal ideal. Agents can disregard an emergent preference on this basis without thinking that they should ultimately repudiate the ranking it recommends.
This reconstruction of a Frankfurtian will-revealing procedure raises basic questions about the credentials of [P.sup.*], [I.sup.*], and [D.sup.*] as necessary conditions for such a procedure. These conditions conflict with the Frankfurt model because they interact in a way that forces one to treat emergent preferences as social choice theorists propose to treat voters: Each is entitled to equal voice and weight in the amalgamation process. But the Frankfurt model reveals difficulties with this assimilation. Consider again the predicament of the tempted spouse. If she believes that [P.sup.*] and [I.sup.*] are appropriate expectations to place on the process by which she determines the content of her will, she presumably will have to identify with the course of infidelity, since this is what her emergent desires unanimously recommend.
Moreover, strictly speaking, considering second-order assessments of the desirability of living in accord with currently emergent desires is ruled out by [I.sup.*]. The reason is that when the tempted spouse invokes her second-order volition as a reason for disregarding her first-order rankings, she is choosing between fidelity and infidelity precisely on the basis of a preference between alternatives that [I.sup.*] deems "irrelevant." [I.sup.*] requires that agents' final orderings be independent of information concerning alternatives not in the set of alternatives under direct consideration. In the Frankfurtian procedure, the tempted spouse reaches a final judgment about fidelity because of the way she ranks two quite different alternatives: a life led in accordance with greed, lust, and vengeance versus a life organized around different motivations. But these two possible modes of life are not themselves members of the set of alternatives under direct consideration, a set that includes only fidelity and i nfidelity on this specific occasion.
It is tempting to object that the tempted spouse's second-order volition to disregard her first-order desires can itself simply be counted as an emergent preference for fidelity over infidelity, on a par with the other first-order emergent preferences. On this supposition, it would no longer be true that the tempted spouse's emergent desires unanimously recommend infidelity over fidelity, since there is now an additional preference for fidelity. This objection misfires in two ways, however. First, stipulating that the second-order volition counts as an emergent preference for fidelity may allow the Frankfurtian procedure to satisfy [P.sup.*], but only by raising a new question about [D.sup.*]. [D.sup.*] requires that there be no criterion or emergent preference dimension such that it determines the final ranking regardless of the rankings recommended by other emergent preferences (see Riker 1982, 118). But if we count the second-order volition as a direct preference for fidelity over infidelity and the tempt ed spouse opts for fidelity over the objections of all her other preferences, this requirement is violated. Given its special importance in the tempted spouse's self-image, the personal ideal of "spouseworthiness" functions in just such a "dictatorial" way in the formation of her will. But in combination with [I.sup.*] and [P.sup.*], [D.sup.*] denies this sort of special privilege to particular preferences. Instead, these conditions hold will formation hostage to a principle of "one desire, one vote." As the tempted spouse illustrates, however, the process of will formation is not necessarily well captured within the terms of this principle. Deliberation about the content of one's will and about how to be true to oneself often is precisely and legitimately about which preferences to privilege over others and about which alternatives are relevant in a particular instance. [P.sup.*], [D.sup.*], and [I.sup.*] take these issues off the deliberative agenda by deciding them in advance according to an arguably inapp ropriate principle of equal consideration.
This hints at the second way in which the objection misfires. The objection simply stipulates that we assimilate second-order volitions and first-order emergent rankings of the alternatives. It may well be possible to retrieve a rationale for the [TUPID.sup.*] conditions once we make this stipulation, but it begs the critical question of whether we must view second-order volitions in this way. If this stipulation is independently implausible (because it fails to detect relevant differences between these two kinds of emergent preference), then so much the worse for the [TUPID.sup.*] model of the will. And the stipulation is implausible for this reason. Second-order volitions cannot simply be assimilated without qualification to other first-order rankings of the alternatives. A second-order volition not to be moved to X (where Y is some alternative) when prompted by first-order desires A, B, and C is not adequately described as a preference for Y over X. There may be other first-order desires (P, Q, R) that, i f emergent, also would recommend X but to which the operative second-order volition has, as it were, no objection. This is hard to see in the tempted spouse example, because it is difficult to imagine "spouseworthy" motivations for infidelity, but not in other cases. John may choose not to date Sarah because he thinks his desires for doing so are suspect (lust, a desire to get back at her current husband, wanting to access Sarah's millions), but he remains open to the possibility that other desires for going out with Sarah (e.g., true love), if they emerge, could be allowed to determine his will.
This does not mean that the tempted spouse's eventual will cannot be seen as expressing a certain kind of desire to refrain from infidelity. That description remains available, subject to an important qualification: The tempted spouse's eventual desire to refrain from infidelity is the result of a deliberative assessment of her emergent desires on this occasion, not one of the initially emergent desires for or against fidelity. Appealing to the ideal of spouseworthiness and associated second-order volitions is a method for arriving at a final judgment, and the outcome of this appeal can legitimately be described as a desire to refrain from infidelity. But this does not mean that in the first instance the second-order desire is itself a desire for or against fidelity. Rather, it is essentially a desire not to be the kind of person who acts on the sorts of motivations implicated in the first-order rankings of fidelity and infidelity.
This is not to deny that the desire to live up to the operative ideal is itself an emergent desire, but it only ranks fidelity over infidelity indirectly, by rejecting the motives on which the initially unanimous first-order recommendation is based on this occasion. Assessing the relation between one's currently emergent first-order rankings and some privileged personal ideal is surely a perfectly legitimate way to process one's initially emergent preferences into a final judgment about one's will on a given occasion. As long as we take [P.sup.*], [I.sup.*], and [D.sup.*] seriously, however, we must regard such a procedure with suspicion, for these conditions prohibit agents from assigning any greater weight to second-order volitions than to any other ranking. Yet, flattening out the will in this fashion arguably misrepresents the special character and legitimate role of privileged second-order volitions in the formation of the will. Substantive doubts of this sort cannot be removed simply by stipulating a p riori that this compressed model of the will is the correct one.
RECONSTRUCTING THE COLLECTIVE WILL
I have called into question the appropriateness of [P.sup.*], [I.sup.*], and [D.sup.*] as conditions under which individual agents determine the content of their will, given emergent desires or criterial rankings of the alternatives under consideration. Can the lessons of this analysis be extrapolated to the case of collective self-government and, in particular, democratic self-government? At first sight, this seems unlikely, because I have claimed that it is problematic to treat agents' emergent preferences as if they were voters expressing judgments entitled to equal consideration. But this is not so problematic in the collective case, in which we are dealing precisely with voters' judgments, and to which the democratic principle of equal consideration seems entirely appropriate. Perhaps the argument I have given simply shows that the individual psyche is typically not organized according to democratic principles. But if a democratic polity is just a group of individuals whose own views about public policy must be weighted fairly, then a presumption in favor of equal consideration may seem to follow very naturally, and with it a cogent rationale for P, D, and I. It is because these conditions seem to express a norm of democratic reflexive authenticity that social choice theory is so disturbing to conceptions of democratic self-government.
I will now show how the Frankfurtian argument offered above helps us see why F, D, and I may not be essential requirements for any reasonable standard of democratic authenticity. The discussion of second order volitions questioned the requirement that the individual will depend exclusively on the profile of currently emergent preferences without any assessment of their character in relation to operative ideals. Does this proviso carry over to the collective case?
Consider the case of a jury in a criminal trial, as that institution is currently understood in common law jurisdictions. Suppose that initially the jurors are unanimously in favor of conviction, but upon deliberation it emerges that this unanimity is largely an artifact of prejudice, tendentious interpretation of the evidence, and disregard for the reasonable doubt standard. Under these (hardly exotic) conditions, the jurors might reasonably decide that to permit their joint will to be determined by these initially emergent preferences would fall short of the ideal of juryhood that defines their association. In that sense, the verdict implied in the initial consensus could not represent their will qua jury. But at this point the jurors need not conclude that they will eventually acquit the defendant. Their rejection of the original recommendation to convict was not based on a direct judgment of the proper verdict but on an indirect assessment of the way it was reached. No definitive conclusion about the pro per verdict can be inferred at this stage. The jury simply resolves to continue the discussion until it reaches a verdict on terms compatible with a certain ideal of how juries ought to make decisions. Suppose that eventually it votes unanimously for acquittal after a sincere effort to apply the reasonable doubt test.
Keeping this example in mind, suppose that we must determine rules and conditions under which the votes of a jury can genuinely be described as expressing a jury's will. Must we accept F, D, and I as appropriate conditions? If so, we could not make any clear distinction between the two votes reached by the jury in the example. Both the initial vote for conviction and the later vote for acquittal are unanimous, and accepting either is compatible with P, D, and I. From the viewpoint of the TUPID model, the only salient difference between the two votes is the time at which they occurred and the fact that at those moments the pattern of emergent preferences was different. But P, D, and I by themselves give us no basis on which to prefer one to the other, apart from the Hobbesian ground that the later unanimity immediately preceded the moment of resolution and so was contingently decisive. But the jury need not have changed its mind. Had it not done so, by the lights of P, D, and I, the earlier decision could hav e stood equally well as the jury's will.
It is not obvious that this is the only conceivable or reasonable standard of reflexive authenticity for juries. From another point of view the first unanimous verdict represents the jury's will less successfully than the second, but this perspective will elude us as long as we think of a jury as nothing more than a collection of separate individuals free to use whatever criteria they wish to determine their own view of the appropriate verdict. If this is what a jury is, then P, D, and I will seem perfectly appropriate, and the failure of these conditions to discriminate between the two verdicts becomes unproblematic. For on this understanding, a jury is simply an empty vessel whose character is determined entirely by the preferences of the individual jurors who happen to make it up at a given moment. Since both unanimous verdicts in the example represent them equally well (at different times), there is no reason to expect that one verdict represents the jury's will more successfully than the other.
Clearly, this is not the only available conception of a jury. Another defines it as a form of association that demands that its members be moved by certain considerations rather than others. A group could completely disregard the appropriate considerations and perhaps reach joint decisions, but it could not reasonably regard them as reflecting a will that is theirs qua jury. It would be basing decisions on considerations that are, in effect, unworthy of a jury. Indeed, in a sense they are failing to be a jury, or at least failing to behave like one. Adopting this more restrictive conception does not in any way show that the ideal of juryhood is itself worthy; jurors might have their own excellent reasons for wanting to subvert conventional legal practice. But as long as that ideal is partly definitive of a jury (for good or ill), a standard of reflexive authenticity for juries can reasonably demand that jurors take that ideal seriously. Moreover, it seems clear that the jury in the example fulfills this expe ctation. When they rejected the initial verdict, jurors did so because they reflected upon their initially emergent preferences and found their character is inappropriate given their self-image as a jury. Insofar as they aspired to juryhood, they had to conclude that their first verdict could never count as their will and must be revised, even though unanimous. Although P, D, and I do not disallow this revision, they also do not require it; these conditions could allow the first verdict to count as the jury's will. But a standard of reflexive authenticity based on the alternative conception of juryhood might well require the jurors to disallow the first verdict, so that alternative view must exclude P, D, and I.  It follows that if this alternative conception is reasonable, then we have found at least one collective will-revealing procedure in which relaxing the TUPID conditions is innocent.
One might object that there are fundamental differences between a jury and the sovereign electorate of a purportedly self-governing democracy. There is some truth to this: Directing democratic debate about public policy along the narrow lines defined by a judge who instructs a jury to disregard all criteria apart from the reasonable doubt standard would surely impinge unduly upon democratic freedoms. Moreover, in the jury example constraints are imposed externally by the judge and the wider legal apparatus in which the jurors find themselves immersed, but the idea of democratic self-government seems to require that any relevant restrictions be imposed from within.
The force of these points is easily exaggerated, however, particularly if we draw an unnaturally sharp line between procedure and substance in the context of democratic practice. Stressing the way in which democracy asserts the priority of certain decision procedures over substantive political viewpoints can obscure the way in which democratic procedure can itself function as a substantive social ideal. That social ideal need not specify any particular view about the goals of public policy. But, as in the jury example, it unavoidably defines a certain ideal about the terms on which certain policy alternatives are to be identified as the will of a democratic public. To the extent that citizens represent themselves as members of a collectivity with the distinctive character of a democratic public, they may reasonably impose restrictions on the range of first-order values and motivations that could in principle determine their joint will. Moreover, the justification for such restrictions need not be based upon an elaborated and defensible theory of the public's best interests, or of the common good. To establish that such a democratic public is self-governing, it would be necessary only to justify restrictions on the following more slender ground: They are a necessary condition for plausibly ascribing to the authors of the social will the distinctive characteristics of a democratic public."  If citizens see themselves as members of such a public, they would be untrue to their collective character if they were to ignore such restrictions.
These restrictions will vary according to the content of the relevant democratic ideal, but it is not difficult to imagine likely candidates. Excluded motivations might include, minimally, and formally: determining votes by all consulting the same oracle and following its directives or by tossing a coin. More substantively, they might include ranking alternatives entirely on the basis of judgments about irrelevant features of the people who advocate them (e.g., skin color, ethnic origin, and religious affiliation) and determining one's vote on the basis of criteria that give absolutely no weight to, or completely discount, the values of free and open debate, compromise, accountability, equal consideration, and transparency. These restrictions neither amount to nor necessarily rest on a particular view of the common good or the substantive goals of public policy, but a society that ignores these desiderata clearly fails to meet expectations implicit in the idea of democratic agency. As in the jury example, citizens might reasonably conclude that when a society acts on the basis of such motivations and values it fails to live up to the ideal of an open, inclusive democratic society.  Unlike the jury example, it is not obvious that these standards are necessarily "externally imposed" upon citizens. They simply may be embedded deeply in the institutional history and discursive practice of a given polity, such that it would seem very unnatura l to regard citizens who apply them as appealing to anything alien or extraneous. 
The central point is that practices of compromise, open debate and inclusion, and the associated deliberative commitments to values such as fairness, consideration, and freedom are partly constitutive of democratic agency.  To the extent that restrictions essential to these practices fail to operate, one may legitimately doubt whether the results of the process can qualify as the will of a group with the distinctive character of a democratic public. By contrast, P, D, and I involve an interpretation of democratic authenticity that pays no attention to the character of the motivations behind individual votes and preferences. The only information that is allowed to count concerns patterns in the list of individual rankings offered by citizens. This is the same "empty vessel" interpretation of reflexive authenticity discussed earlier: The character of a democratic public can be whatever the preferences of its members make it. From a Frankurtian viewpoint, democratic authenticity requires second-order restri ctions on the sorts of motivations and values upon which voters' judgments are based. The standard objection to such restrictions is that when they are applied, "social choice is based as much on the restriction[s] as it is on individual judgments" (Riker 1982, 117). The response is to insist that although such restrictions may not be based on individual judgments on specific occasions, observing them may nevertheless be a necessary condition for certifying that the individuals expressing those judgments comprise a group with the distinctive character of a democratic public. The criteria for attributing to a collectivity a democratic character automatically entail various restrictions on the range of motivations allowed to constitute its will. Once we see this, we can dismiss the charge that such restrictions necessarily preempt democratic authenticity because they are "imposed upon" a democratic public as if from outside. Frankfurtian restrictions are presupposed in the very concept of a democratic public it self. In a sense, they cannot be imposed; if they are not operative, there simply will be no democratic public on which to impose. That is why they can legitimately be regarded as essential to an appropriate test of reflexive authenticity for democratic publics.
Given this, citizens who think of themselves as members of a democratic public have reasons to reject P, D, and I as criteria of democratic authenticity. Frankfurt's terminology helps illuminate what is at stake: For a democratic public to endorse F, D, and I as its criteria of reflexive authenticity would institutionalize wantonness. In virtue of its accepted self-image, such a public would have a second-order desire to fulfill its aspiration to be genuinely democratic. But the effect of endorsing P, D, and I would be to deny any special privilege to that ideal in the formation of its will. Such an authorization would instantiate wantonness by undermining the basis for any plausible discrimination between the fulfilment of society's aspiration to be democratic and those first-order preferences or emergent desires that might signify its betrayal.
Replacing F, D, and I with a more restrictive criterion of democratic authenticity does not itself guarantee that such betrayals will not occur (although it might imply institutional--perhaps in particular constitutional--mechanisms to disable or impede them). But this does not affect the key theoretical point that from the perspective of an ideal of democratic self-government it is irrational to institutionalize wantonness. Surely, there is a fundamental difference between a society whose aspiration to be genuinely democratic has been betrayed in practice and a society that has institutionalized a license for this kind of betrayal. It is the difference between Frankfurt's wanton and Frankfurt's unwilling addict (or the tempted spouse). From the Frankfurtian perspective, Arrow's theorem becomes much less disturbing to the idea of democratic self government. The claim that the TUPID conditions cannot result in a coherent will can now be interpreted simply as a formalization of the claim that wanton collectivi ties, like wanton individuals, are systematically unable to form a will.
Clearly, these arguments are not intended as a solution to Arrow's theorem on its own terms, if by that one means an all-purpose algorithm for guaranteeing that voting procedures will not result in cyclical preferences. My argument also is not meant as a refutation of the theorem; I do not believe that Arrow's theorem can be refuted. Rather, I have attempted to put the impossibility theorem in proper perspective within the context of self-government. Social choice theorists have made strong claims that Arrow's theorem establishes the logical absurdity of the notion of a popular democratic will. That is only true if the expectations imposed by social choice theory are necessary conditions for any reasonable will-revealing procedure. I have suggested that this assumption is dubious. We may have been looking in the wrong place for democratic will-revealing procedures and tacitly relying on an unreasonably stringent criterion of collective self-government.
The specification and application of my broader understanding of will formation are complex matters and face many difficulties. The result may not be an understanding of the social will with the formal elegance and rigor offered by the social choice model. That would be a real loss. But there is also the potential for compensating gains. A wider notion of the social will may illuminate a number of issues that the social choice model leaves obscure. In closing, I offer three thoughts along these lines.
First, many insights are offered by social choice theorists. I merely claim that the model of will formation presupposed by Arrow and his successors is not uniquely reasonable and appropriate to all choice situations. Furthermore, some choice situations may not be governed by relevant second-order volitions. In such cases, the TUPID conditions may yet be appropriate, and it will follow that there is no way to determine the content of the social will with regard to the specific alternatives under consideration. Conversely, the presence of second-order volitions may cause preferences over alternatives to have the property of single-peakedness (Riker 1982, 128). It has long been understood that where this property exists, the Arrow problem does not arise. Integrating social choice theory into the broader perspective proposed here may allow us to define more sharply the preconditions for the formation of a social will.
Second, the perspective offered here brings into the foreground questions about the interface between second-order volitions that define citizenship and those that operate in other arenas. For example, individuals have personal ideals, and the ideals of democratic self-government also make demands on them. Different personal second-order volitions might threaten or facilitate the organization of a democratic will, and it is important to know which. Moreover, semipublic institutions such as political parties form wills using their own characteristic second-order volitions. Perhaps personal identification with such ideals being a "loyal Republican" or a "faithful trade unionist," and the sanctions imposed by such institutions for disloyalty, may help reduce or increase conflicts between personal ideals and ideals of citizenship. Understanding the role of semipublic institutions might tell us much about the conditions under which a public will can be organized. There is also the related question of the range of motivations affiliated with particular parties or groups that a given ideal of democratic citizenship will tolerate: Unfortunately, the people of Austria and Northern Ireland have recently found themselves embroiled in just this issue. Arguments to the effect that certain kinds of parties, and the values that identify them, are "beyond the pale" or inimical to the conduct of democratic decision making may be essential to the practice of democratic self-government.
Third, the Frankfurtian position preserves the idea, deeply embedded in ordinary usage, that wills, once identified, may nevertheless be betrayed in practice. Drawing on Aristotle, Hurley (1989, 136-7) gives a fascinating illustration of how this idea can have collective applications. Members of a democratic public may "will" an end to racial discrimination (perhaps because they take this to be mandated by democracy itself), yet find that their own will is frequently betrayed in their interactions as private citizens (de facto discrimination, powerful groups using their favored position to circumvent or exploit loopholes in relevant statutes). The line of argument opened here provides a framework for analyzing this kind of phenomenon and can help us perceive a politics of the will in the characteristic discourse that develops around the attempt to impose legal and social controls on citizens' behavior.
These suggestions are sketchy and present serious difficulties, but there is no reason to suppose a priori that these difficulties are impossible to overcome. The social choice theorists suggested that the notion of a collective will is indeed an a priori impossibility. If the argument of this article has succeeded in dispelling these alleged impossibilities and in opening up a viable line of argument, the progress, though modest and incomplete, will nevertheless be real.
Colin Bird is Assistant Professor of Government and Foreign Affairs, University of Virginia, Charlottesville, VA 22903.
The author thanks Talbot Brewer, James Cargile, Dan Devereaux, Joshua Dienstag, Paul Freedman, George Klosko, Charles Kromkowski, Debra Morris, Steven Wall, Brantly Womack, and several anonymous referees for their extremely helpful comments on earlier versions of this article.
(1.) This sort of claim is normally brought in under the aegis of claims about the virtues of citizenship. Such virtues require that citizens be motivated by a concern for the public good rather than by their self-interest (their "particular wills," in Rousseau's terminology). Self-government, in this view, requires a virtuous citizenry. For a recent statement of the idea, see Barber 1993, 66.
(2.) For a superb illustration of this tendency to dissociate democracy and self-government in modern empirical research, see Duverger 1954, 424: "No people has ever been known to govern itself and none ever will. All government is oligarchic: it necessarily implies the domination of the many by the few.... All government implies discipline ... by definition constraint is external to the constrained. A people does not constrain itself, it is constrained. It does not govern itself, it is governed."
(3.) Note, however, that this is perfectly compatible with the claim that any authentically good society would have to be to a high degree self-governing.
(4.) Arrow ( 1963, 85) had already anticipated this proposal in his original monograph, suggesting that it "has much in common with the statistical problem of pooling the opinions of a group of experts to arrive at a best judgment; here individuals are considered experts at detecting the moral imperative." If one assumes the existence of a "right answer," and individuals are likely to arrive at the correct decision on their own more than 50% of the time, then the probability that a majority will be right increases as the size of the relevant group increases. See Barry 1990, 292-3.
(5.) Just this line of argument has led a host of critics to demur at Rousseau's notorious claim about forcing citizens to be free. They think that Rousseau gave the game away when he claimed that the "General Will is never wrong" and that popular participation in expressing it causes "the public to learn to know what it wants." Social choice theory tends to reinforce the suspicion that Rousseau's democratic and voluntarist terminology disguises an essentially Platonic dogmatism.
(6.) These conditions are taken from Hurley 1989, 229.
(7.) The point here is precisely not that this results in an indefensible conception of the will: Many philosophers and theorists have defended views of the will that seem consistent with this one. Hobbes, for example, can be interpreted as suggesting that individuals are only able to arrive at decisions at all by violating the nondictatorship condition. In Hobbes's view, deliberation is nothing other than the process by which a sequence of values (embodied by a series of emergent appetites and aversions) compete for dominance within the mind of an agent. An individual finally wills an action when one or other of these competing values wins out. In this way, the victorious appetite can be said to dictate the decision. That "dictated" decision represents the individual's will, just as the arbitrary decisions of the sovereign represent the will of the collectivity after the social contract has instituted a Leviathan. So it is perfectly possible to defend a notion of the will that violates one or other of Arrow' s conditions. On this point generally, see Mayo 1956.
(8.) I am not the first to suggest this general strategy of analysis. Hurley (1989, 234-41) explores a similar line of argument. But I find her argument unsatisfactory. Hurley accepts [P.sup.*] and [D.sup.*] as appropriate conditions on individual deliberation but rejects [U.sup.*] and [I.sup.*]. To my mind, [U.sup.*] is the one condition that does make sense in the individual case (see Steedman and Krause 1986, 208), so my own arguments will target [P.sup.*], [I.sup.*], and [D.sup.*].
(9.) It seems possible in principle that second-order volitions may be capable of producing, or causing one to have, particular first-order desires. See, for example, Raz 1986, 32-3.
(10.) Suppose, for example, that after the first unanimous vote, the jury holds a second vote on whether to discuss the verdict further or let the recommendation stand. In this second vote, all but one agree to let the conviction stand, because they want their jury duty to be over and done with. In our example the lone holdout is right to think that the initial vote, albeit unanimous, may not represent a will that is theirs qua jury, so a will-revealing procedure based on our alternative understanding of juryhood might require that his single vote determine their will as a jury at that moment--in this case to continue deliberation. But this seems to require relaxing D.
(11.) Or they are characteristic of some polity that aspires to some more substantive social ideal that is nested within and compatible with democratic ideals and procedures (e.g., "the American way").
(12.) The relevant restrictions would therefore be what Goodin (1986, 78) helpfully calls "input filters."
(13.) This may imply a certain conservative bias, in the sense that embedded social ideals (such as democracy) require citizens to acknowledge an overriding allegiance to the extant practices, cultures, and institutions in which such ideals are latent. But perhaps that is necessary in order for a political group to have a persisting identity over time such that it can be in principle the subject of its own destiny. This tends to vindicate the Aristotelian idea that a society's constitution is both the source of a local consensus on certain social values and a political society's essential identifying attribute (Aristotle 1981, 176). Perhaps the real lesson of the impossibility theorems is that a social will cannot emerge in a social vacuum: There must be at least some nonarbitrary constraints informing citizens' joint self-understanding as a society organized around certain ideals and values. Only then can they see themselves as participants in an overriding political identity to which their joint decisions a nd actions may (or may not) be loyal. But note that constraints of this kind are crucially different from those imposed by the various forms of the General Will argument we considered earlier. Under that argument, the constraints are imposed by some independently authoritative account of the common good so as to increase the likelihood that the right answer will emerge from appropriately designed decision procedures.
(14.) This does not mean that democratic publics must consciously take these values as their concrete goals, only that consciously or not, these restrictions must in fact shape the character of the political process. The property of acting like a democratic public may, for example, be what Elster (1986, 122-7) calls a by-product. This highlights another difference between the line of argument developed here and epistemic populism. In the latter view, citizens are required to embrace consciously a particular collective goal (some substantive view of the common good) as part of the political process. In my view, this need not be so.
Aristotle 1981. The Politics. London: Penguin
Arrow, Kenneth J.  1963. Social Choice and Individual Values. New Haven, CT: Yale University Press.
Barber, Benjamin R. 1993. "Reductionist Political Science and Democracy." In Reconsidering the Democratic Public, ed. George E. Marcus and Russell L. Hanson. University Park: Penn State University Press. Pp. 65-72.
Barry, Brian. 1990. Political Argument. Hemel Hempstead: Harvester Wheatsheaf.
Cohen, Joshua. 1986. "An Epistemic Conception of Democracy." Ethics 97 (October): 26-38.
Coleman, Jules, and John Ferejohn. 1986. "Democracy and Social Choice." Ethics 97 (October): 6-25.
Duverger, Maurice. 1954. Political Parties, trans. Barbara and Robert North. London: Methuen.
Elster, Jon. 1986. "The Market and the Forum: Three Varieties of Political Theory." In Foundations of Social Choice Theory, ed. Jon Elster and Aanund Hylland. Cambridge: Cambridge University Press. Pp. 103-33.
Frankfurt, Harry. 1988. The Importance of What We Care About. Cambridge: Cambridge University Press.
Goodin, Robert. 1986. "Laundering Preferences." In Foundations of Social Choice Theory, ed. Jon Elster and Aanund Hylland. Cambridge: Cambridge University Press. Pp. 75-101.
Grofman, Bernard, and Scott Feld. 1988. "Rousseau's General Will: A Condorcetian Perspective." American Political Science Review 82 (June): 567-76.
Hurley, Susan. 1989. Natural Reasons: Personality and Polity. Oxford: Oxford University Press.
May, Kenneth. 1954. "Intransitivity, Utility, and the Aggregation of Preference Patterns." Econometrica 22 (January): 1-13.
Mayo, Bernard. 1956. "Is There a Case for the General Will." In Philosophy, Politics and Society, ed. Peter Laslett. New York: MacMillan. Pp. 92-8.
McClean, Iain. 1991. "Rational Choice and Politics." Political Studies 39 (September): 496-512.
McKelvey, Richard. 1979. "General Conditions for Global Intransitivities in Formal Voting Models." Econometrica 47 (September): 1085-111.
Michels, Robert. 1916. Political Parties: A Sociological Study of the Oligarchical Tendencies of Modem Democracy. London: Jarrold.
Plott, Charles. 1976. "Axiomatic Social Choice Theory: An Overview and Interpretation." American Journal of Political Science 20 (August): 511-96.
Radcliff, Benjamin. 1992. "The General Will and Social Choice Theory." The Review of Politics 54 (Winter): 34-49.
Raz, Joseph. 1986. The Morality of Freedom. Oxford: Oxford University Press.
Riker, William. 1982. Liberalism against Populism: A Confrontation between the Theory of Democracy and the Theory of Social Choice. San Francisco: W. H. Freeman.
Roemer, John. 1996. Theories of Distributive Justice. Cambridge, MA: Harvard University Press.
Schelling, Thomas. 1984. Choice and Consequence. Cambridge, MA: Harvard University Press.
Schofield, Norman. 1985. Social Choice and Democracy. Berlin: Springer-Verlag.
Schumpeter, Joseph. 1942. Capitalism, Socialism, and Democracy. New York: Harper & Row.
Sen, Amartya. 1970. Collective Choice and Social Welfare. San Francisco: Holden-Day.
Steedman, Ian, and Ulrich Krause. 1986. "Goethe's Faust, Arrow's Possibility Theorem and the Individual Decision-Taker." In The Multiple Self, ed. Jon Elster. Cambridge: Cambridge University Press. Pp. 197-233.
|Printer friendly Cite/link Email Feedback|
|Publication:||American Political Science Review|
|Date:||Sep 1, 2000|
|Previous Article:||The Sovereignless State and Locke's Language of Obligation.|
|Next Article:||The Fear of Death and the Longing for Immortality: Hobbes and Thucydides on Human Nature and the Problem of Anarchy.|